OpenAI Python SDK v3.41.0
OpenAI Python SDK v3.41.0 adds typed references for Realtime responses, tool usage models, agent session filtering by metadata, and WebSocket binary frame recovery after reconnects.
Why this signal matters
The OpenAI Python SDK release v3.41.0 introduces several feature additions and bug fixes. Key features include support for typed references in Realtime response inputs ( 4189), the addition of response tool usage models ( 4201), the ability to filter agent sessions by metadata ( 4200), and automatic recovery of queued binary frames following WebSocket reconnections ( 4191). Bug fixes address detecting incomplete WebSocket responses ( 4187), handling interruptions and disconnects in push-to-talk examples ( 4192), and keeping consumer fixtures clean under strict Pyright configurations ( 4204). Chores include documentation updates regarding upcoming model shutdowns and image models, generated-file ownership enforcement, and migrating development type checking to ty.
Actionable summary
OpenAI Python SDK v3.41.0 introduces typed references in Realtime response inputs, response tool usage models, metadata-based agent session filtering, and WebSocket queued binary frame recovery upon reconnection. It also includes fixes for incomplete WebSocket response detection and push-to-talk disconnections.
- Agent usefulness
- 77/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Upgrade openai Python SDK to v3.41.0 if using Realtime APIs or agent session metadata filtering.
- Review upcoming model shutdown documentation to ensure deprecated models are not in active use.
- Update push-to-talk or WebSocket implementations to leverage improved reconnection frame recovery and disconnect handling.